Text To Image Synthesis

This is an experimental tensorflow implementation of synthesizing images. The images are synthesized using the GAN-CLS Algorithm from the paper Generative Adversarial Text-to-Image Synthesis. This implementation is built on top of the excellent DCGAN in Tensorflow. The following is the model architecture.

Image Source : Generative Adversarial Text-to-Image Synthesis Paper

Requirements

• TensorLayer 1.3.1 (or later)
• Tensorflow 0.10 (or later)
• NLTK : for tokenizer

Datasets

• The model is currently trained on the flowers dataset. Download the images from here and save them in 102flowers/102flowers/*.jpg. Also download the captions from this link. Extract the archive, copy the text_c10 folder and paste it in 102flowers/text_c10/class_*.

Codes

• train_dcgan.py use DCGAN to generate new images.
• train_txt2im.py use GAN-CLS to generate new images conditioned on text.
• utils.py helper functions.

References

• Generative Adversarial Text-to-Image Synthesis Paper
• Generative Adversarial Text-to-Image Synthesis Torch Code
• Skip Thought Vectors Paper
• Skip Thought Vectors Code
• Generative Adversarial Text-to-Image Synthesis with Skip Thought Vectors TensorFlow code
• DCGAN in Tensorflow

Credits

License

Apache 2.0